Clustering Model - Credit Card Dataset¶

Data Loading and Cleaning¶

In [1]:
# Loading libraries

import pandas as pd
import numpy as np
import seaborn as sns # For creating the plots
In [2]:
# Creating a data frame using the csv file
df_original = pd.read_csv('credit_dataset.csv')
df_original.head()
Out[2]:
CUST_ID BALANCE BALANCE_FREQUENCY PURCHASES ONEOFF_PURCHASES INSTALLMENTS_PURCHASES CASH_ADVANCE PURCHASES_FREQUENCY ONEOFF_PURCHASES_FREQUENCY PURCHASES_INSTALLMENTS_FREQUENCY CASH_ADVANCE_FREQUENCY CASH_ADVANCE_TRX PURCHASES_TRX CREDIT_LIMIT PAYMENTS MINIMUM_PAYMENTS PRC_FULL_PAYMENT TENURE
0 C10001 40.900749 0.818182 95.40 0.00 95.4 0.000000 0.166667 0.000000 0.083333 0.000000 0 2 1000.0 201.802084 139.509787 0.000000 12
1 C10002 3202.467416 0.909091 0.00 0.00 0.0 6442.945483 0.000000 0.000000 0.000000 0.250000 4 0 7000.0 4103.032597 1072.340217 0.222222 12
2 C10003 2495.148862 1.000000 773.17 773.17 0.0 0.000000 1.000000 1.000000 0.000000 0.000000 0 12 7500.0 622.066742 627.284787 0.000000 12
3 C10004 1666.670542 0.636364 1499.00 1499.00 0.0 205.788017 0.083333 0.083333 0.000000 0.083333 1 1 7500.0 0.000000 NaN 0.000000 12
4 C10005 817.714335 1.000000 16.00 16.00 0.0 0.000000 0.083333 0.083333 0.000000 0.000000 0 1 1200.0 678.334763 244.791237 0.000000 12
In [3]:
# The number of rows and columns in the dataset
print('The number of Rows in the dataset:' , df_original.shape[0])
print('The number of Columns in the dataset:', df_original.shape[1])

The number of Rows in the dataset: 8950
The number of Columns in the dataset: 18
In [4]:
# Dropping the first column since it is a primary key and not required for our analysis
df_original.drop('CUST_ID', axis=1, inplace=True)
In [5]:
# Replacing all infinity values with nan values
df_original.replace([np.inf, -np.inf], np.nan, inplace=True)

# Removing all nan values
df_original = df_original.dropna()
In [6]:
# Checking the stats of all the columns in the dataframe
df_original.describe()
Out[6]:
BALANCE BALANCE_FREQUENCY PURCHASES ONEOFF_PURCHASES INSTALLMENTS_PURCHASES CASH_ADVANCE PURCHASES_FREQUENCY ONEOFF_PURCHASES_FREQUENCY PURCHASES_INSTALLMENTS_FREQUENCY CASH_ADVANCE_FREQUENCY CASH_ADVANCE_TRX PURCHASES_TRX CREDIT_LIMIT PAYMENTS MINIMUM_PAYMENTS PRC_FULL_PAYMENT TENURE
count 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000 8636.000000
mean 1601.224893 0.895035 1025.433874 604.901438 420.843533 994.175523 0.496000 0.205909 0.368820 0.137604 3.313918 15.033233 4522.091030 1784.478099 864.304943 0.159304 11.534391
std 2095.571300 0.207697 2167.107984 1684.307803 917.245182 2121.458303 0.401273 0.300054 0.398093 0.201791 6.912506 25.180468 3659.240379 2909.810090 2372.566350 0.296271 1.310984
min 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 50.000000 0.049513 0.019163 0.000000 6.000000
25% 148.095189 0.909091 43.367500 0.000000 0.000000 0.000000 0.083333 0.000000 0.000000 0.000000 0.000000 1.000000 1600.000000 418.559237 169.163545 0.000000 12.000000
50% 916.855459 1.000000 375.405000 44.995000 94.785000 0.000000 0.500000 0.083333 0.166667 0.000000 0.000000 7.000000 3000.000000 896.675701 312.452292 0.000000 12.000000
75% 2105.195853 1.000000 1145.980000 599.100000 484.147500 1132.385490 0.916667 0.333333 0.750000 0.250000 4.000000 18.000000 6500.000000 1951.142090 825.496463 0.166667 12.000000
max 19043.138560 1.000000 49039.570000 40761.250000 22500.000000 47137.211760 1.000000 1.000000 1.000000 1.500000 123.000000 358.000000 30000.000000 50721.483360 76406.207520 1.000000 12.000000
In [7]:
# Applying Normalization to the data 
from sklearn.preprocessing import MinMaxScaler
norm = MinMaxScaler().fit(df_original)
df = norm.transform(df_original)
In [8]:
# Applying Standardization to the data
# from sklearn.preprocessing import StandardScaler  
# scaler = StandardScaler().fit(df)  
# df = scaler.transform(df)

I have chosen Normalization instead of Standardizatoin since its giving lower values of SSE¶

K-Means Clustering Algorithm¶

In [9]:
# Importing the necessary library

from sklearn.cluster import KMeans
In [10]:
# Creating a copy of the dataframe

X = df

Running K-Means with 2 clusters but with differnt seeds to notice the effect on SSE¶

In [11]:
kmeans1_1 = KMeans(n_clusters=2, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans1_1.inertia_ #.inertia_ is the method used to get the SSE
sse1
Out[11]:
3300.7155556099283
In [12]:
kmeans1_2 = KMeans(n_clusters=2, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans1_2.inertia_
sse2
Out[12]:
3300.7148795068624
In [13]:
kmeans1_3 = KMeans(n_clusters=2, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans1_3.inertia_
sse3
Out[13]:
3300.7150608010725
In [14]:
# Recording the minimum SSE
min_sse = min(sse1,sse2,sse3)
min_sse
Out[14]:
3300.7148795068624
In [15]:
# Appending the min SSE to the list called Errors

Errors = []
Errors.append(min_sse)

Running the algorithm for K = 3¶

In [16]:
# Running KMeans for 3 different random states

kmeans2_1 = KMeans(n_clusters=3, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans2_1.inertia_
kmeans2_2 = KMeans(n_clusters=3, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans2_2.inertia_
kmeans2_3 = KMeans(n_clusters=3, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans2_3.inertia_
In [17]:
# Appending the min sse to the Errors list
Errors.append(min(sse1,sse2,sse3))
Errors
Out[17]:
[3300.7148795068624, 2739.005483194418]

Running the algorithm for K = 4¶

In [18]:
# Running KMeans for 3 different random states
kmeans3_1 = KMeans(n_clusters=4, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans3_1.inertia_
kmeans3_2 = KMeans(n_clusters=4, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans3_2.inertia_
kmeans3_3 = KMeans(n_clusters=4, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans3_3.inertia_
In [19]:
# Appending the min sse to the Errors list
Errors.append(min(sse1,sse2,sse3))
Errors
Out[19]:
[3300.7148795068624, 2739.005483194418, 2416.173255161476]

Running the algorithm for K = 5¶

In [20]:
# Running KMeans for 3 different random states
kmeans4_1 = KMeans(n_clusters=5, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans4_1.inertia_
kmeans4_2 = KMeans(n_clusters=5, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans4_2.inertia_
kmeans4_3 = KMeans(n_clusters=5, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans4_3.inertia_
In [21]:
# Appending the min sse to the Errors list
Errors.append(min(sse1,sse2,sse3))
Errors
Out[21]:
[3300.7148795068624, 2739.005483194418, 2416.173255161476, 2143.146153164728]

Running the algorithm for K = 6¶

In [22]:
# Running KMeans for 3 different random states
kmeans5_1 = KMeans(n_clusters=6, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans5_1.inertia_
kmeans5_2 = KMeans(n_clusters=6, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans5_2.inertia_
kmeans5_3 = KMeans(n_clusters=6, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans5_3.inertia_
In [23]:
# Appending the min sse to the Errors list
Errors.append(min(sse1,sse2,sse3))
Errors
Out[23]:
[3300.7148795068624,
 2739.005483194418,
 2416.173255161476,
 2143.146153164728,
 1943.9889545565957]

Running the algorithm for K = 7¶

In [24]:
# Running KMeans for 3 different random states
kmeans6_1 = KMeans(n_clusters=7, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans6_1.inertia_
kmeans6_2 = KMeans(n_clusters=7, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans6_2.inertia_
kmeans6_3 = KMeans(n_clusters=7, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans6_3.inertia_
In [25]:
# Appending the min sse to the Errors list
Errors.append(min(sse1,sse2,sse3))
Errors
Out[25]:
[3300.7148795068624,
 2739.005483194418,
 2416.173255161476,
 2143.146153164728,
 1943.9889545565957,
 1755.110505502886]

Running the algorithm for K = 8¶

In [26]:
# Running KMeans for 3 different random states
kmeans7_1 = KMeans(n_clusters=8, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans7_1.inertia_
kmeans7_2 = KMeans(n_clusters=8, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans7_2.inertia_
kmeans7_3 = KMeans(n_clusters=8, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans7_3.inertia_
In [27]:
# Appending the min sse to the Errors list
Errors.append(min(sse1,sse2,sse3))
Errors
Out[27]:
[3300.7148795068624,
 2739.005483194418,
 2416.173255161476,
 2143.146153164728,
 1943.9889545565957,
 1755.110505502886,
 1600.6976005032914]

Running the algorithm for K = 9¶

In [28]:
# Running KMeans for 3 different random states
kmeans8_1 = KMeans(n_clusters=9, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans8_1.inertia_
kmeans8_2 = KMeans(n_clusters=9, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans8_2.inertia_
kmeans8_3 = KMeans(n_clusters=9, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans8_3.inertia_
In [29]:
# Appending the min sse to the Errors list
Errors.append(min(sse1,sse2,sse3))
Errors
Out[29]:
[3300.7148795068624,
 2739.005483194418,
 2416.173255161476,
 2143.146153164728,
 1943.9889545565957,
 1755.110505502886,
 1600.6976005032914,
 1472.0955166546703]

Running the algorithm for K = 10¶

In [30]:
# Running KMeans for 3 different random states
kmeans9_1 = KMeans(n_clusters=10, random_state=99, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse1 = kmeans9_1.inertia_
kmeans9_2 = KMeans(n_clusters=10, random_state=1, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse2 = kmeans9_2.inertia_
kmeans9_3 = KMeans(n_clusters=10, random_state=20, algorithm='elkan',n_init = 10).fit(X) # applying clustering algorithm
sse3 = kmeans9_3.inertia_
In [31]:
# Appending the min sse to the Errors list
Errors.append(min(sse1,sse2,sse3))
Errors
Out[31]:
[3300.7148795068624,
 2739.005483194418,
 2416.173255161476,
 2143.146153164728,
 1943.9889545565957,
 1755.110505502886,
 1600.6976005032914,
 1472.0955166546703,
 1358.663933545895]

Creating a line chart using the errors¶

In [32]:
import matplotlib.pyplot as plt

x_axis = [2, 3, 4, 5, 6, 7, 8, 9, 10]
y_axis = Errors

plt.plot(x_axis, y_axis,'-or')
plt.title('Line Chart')
plt.xlabel('Value of K')
plt.ylabel('Corresponding SSE')
plt.show()

According to elbow method the best cluster number is 5¶

In [ ]:

Cluster Visualizations¶

Here I'm using umap package to reduce dimensions of the data to two dimensions so that I can visualize the data.

In [33]:
# Installing the package
%pip install umap-learn

Requirement already satisfied: umap-learn in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (0.5.4)
Requirement already satisfied: numpy>=1.17 in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from umap-learn) (1.24.3)
Requirement already satisfied: scipy>=1.3.1 in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from umap-learn) (1.10.1)
Requirement already satisfied: scikit-learn>=0.22 in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from umap-learn) (1.2.2)
Requirement already satisfied: numba>=0.51.2 in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from umap-learn) (0.57.0)
Requirement already satisfied: pynndescent>=0.5 in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from umap-learn) (0.5.10)
Requirement already satisfied: tqdm in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from umap-learn) (4.65.0)
Requirement already satisfied: llvmlite<0.41,>=0.40.0dev0 in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from numba>=0.51.2->umap-learn) (0.40.0)
Requirement already satisfied: joblib>=0.11 in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from pynndescent>=0.5->umap-learn) (1.2.0)
Requirement already satisfied: threadpoolctl>=2.0.0 in /Users/tanishqduttkaushik/anaconda3/lib/python3.11/site-packages (from scikit-learn>=0.22->umap-learn) (2.2.0)
Note: you may need to restart the kernel to use updated packages.
In [34]:
# Importing the umap module and creating a reducer object
import umap
reducer = umap.UMAP()
In [35]:
# Transforming each data to a 2-D point 
embedding = reducer.fit_transform(X)
embedding.shape
Out[35]:
(8636, 2)
In [36]:
# Checking the 2D array
len(embedding)
Out[36]:
8636
In [37]:
# Retrieving the cluster labels

df_original['labels_1'] = kmeans1_1.labels_
df_original['labels_2'] = kmeans2_1.labels_
df_original['labels_3'] = kmeans3_1.labels_
df_original['labels_4'] = kmeans4_1.labels_
df_original['labels_5'] = kmeans5_1.labels_
df_original['labels_6'] = kmeans6_1.labels_
df_original['labels_7'] = kmeans7_1.labels_
df_original['labels_8'] = kmeans8_1.labels_
df_original['labels_9'] = kmeans9_1.labels_
In [38]:
# Plot for 2 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_1]
    )
Out[38]:
<matplotlib.collections.PathCollection at 0x168333ed0>
In [39]:
# Plot for 3 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_2]
    )
Out[39]:
<matplotlib.collections.PathCollection at 0x1664c1910>
In [40]:
# Plot for 4 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_3]
    )
Out[40]:
<matplotlib.collections.PathCollection at 0x1636cb7d0>
In [41]:
# Plot for 5 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_4]
    )
Out[41]:
<matplotlib.collections.PathCollection at 0x166074bd0>
In [42]:
# Plot for 6 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_5]
    )
Out[42]:
<matplotlib.collections.PathCollection at 0x1686722d0>
In [43]:
# Plot for 7 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_6]
    )
Out[43]:
<matplotlib.collections.PathCollection at 0x1667e31d0>
In [44]:
# Plot for 8 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_7]
    )
Out[44]:
<matplotlib.collections.PathCollection at 0x16865a8d0>
In [45]:
# Plot for 9 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_8]
    )
Out[45]:
<matplotlib.collections.PathCollection at 0x1684631d0>
In [46]:
# Plot for 10 clusters

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(style='white', context='notebook', rc={'figure.figsize':(20,10)})

fig, ax = plt.subplots(1,1)


ax.scatter(
    embedding[:, 0],
    embedding[:, 1],
    c=[sns.color_palette()[x] for x in df_original.labels_9]
    )
Out[46]:
<matplotlib.collections.PathCollection at 0x168655050>